StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (New Staff Post, Please read)

[mikejs78]

we definitely do look at things with skepticism. some things are harder to fake than others. and it's not like fakers haven't gone above and beyond before, but at that point, being fooled is a testament to quality of fake rather than people blindly believing things

Was not my intent to accuse anyone here of blindly believing things- it just occurred to me that it seemed like the motherboard was being accepted as fact, and I was curious as to why that was as it's very hard when you come upon a new thread for the first time that is over 3,900 pages long to be able to read every piece of it. Please excuse this newbie just trying to understand . The explanation about the part numbers, matching shipping records makes perfect sense.

 

[Steve]

Do we know the frame rate of GTA6? Switch 2 can probably "easily" run current gen games with a 60fps mode at 30fps (and 540/720 internally). If GTA6 can only run at 30 on PS5 a Switch 2 port might be difficult or even infeasible.

The Witcher 3 is a 30fps on the PS4 and it arrived on Switch.

Plus we still don't know the limitations GTA 6 brings into the table, when we still haven't seen any gameplay

we do know it's arriving on the Series S, so we might have an idea with that?

 

[darthdiablo]

Do we know the frame rate of GTA6? Switch 2 can probably "easily" run current gen games with a 60fps mode at 30fps (and 540/720 internally). If GTA6 can only run at 30 on PS5 a Switch 2 port might be difficult or even infeasible.

No, that's why it's often said it's too early to say. However, if Xbox Series S do get GTA6, unless the game is severely CPU-bound, I think it'd be reasonable to say a Switch 2 port is possible.

 

[Deleted member 511]

The engine/pipeline and input buffers don't matter at all to DLSS execution time

This is false and the documentation explicitly contradicts this claim.

EDIT:

If this is an attempt to bait me into posting the documentation here, it's not going to work.

 

[gabeplays]

Regarding Ampere specifically, it's a 2020 GPU architecture, same as RDNA 2. Current-gen.
PS4 GPU is Bonaire, GCN 1.1, from ~2013. Seven year gap.
Seven years brings with it a lot of refinements, advancements, and efficiency features.
Dedicated hardware for tensor and RT cores as well as the APIs to target them.
Miniaturized tech like NS2 benefits heavily from this.
Deliberately target lower pixel counts, pack in detail, upscale to higher resolutions with AI. Clever clever.
Never mind the gap in CPU, RAM, etc.

The folks who interpret 'PS4-level' as producing 1:1 PS4 results docked will be in for a surprise.

That's why it's honestly fun to see all the low ball estimations. It's fine to be skeptical, but diminishing returns are hitting really hard and so much has changed since the Switch 1 came out. Even if the the Switch 2 was truly nothing more than a handheld PS4, that's still a meaty upgrade over the OG Switch and we know the successor is way, way, more than that.

Games are just becoming so much more homogenous and it feels like the needle has hardly moved in the past decade. When I look at games like God of War, Detroit Become Human, the Uncharted series, Last of Us, and so-on; they're still drop dead gorgeous after all this time. The Switch 2 can handle all of them.

At this junction, it's not about the hardware at all. It's about the programming, optimization, and the business decision if it's worth it to port something over. Given that this is the successor to the third best selling video console in history, I'm not too worried about the future.

If Nintendo could get TOTK and it's incredibly indepth physics to run on the OG Switch, the new hardware should make anything possible.

 

[LiC]

This is false and the documentation explicitly contradicts this claim.

The documentation's example of this is about "engine" memory management, by which they mean the callbacks you provide to DLSS for it to be able to allocate VRAM. If you give it a callback that has suboptimal performance, or just one with a sleep(5000) in it, then yes, that will affect the time to get your upscaled color buffer back, because it has to allocate some VRAM before it can do that. But that has nothing to do with the execution time of the actual upscaling operation, which, again, is what people are always talking about here.

EDIT:

If this is an attempt to bait me into posting the documentation here, it's not going to work.

Erm, the DLSS documentation is available on Github.

 

[Danny]

IT. IS. JUST. A. PS5. GAME.
There's nothing special about it. The Switch has the hardware to easily run it.

And the funny thing is that they will likely do as every other publisher, rush to make it Steam deck verified later, funny how all those seemingly impossible to port next gen AAA games all rush to steam deck verify their games afterwards.

 

[Deleted member 511]

The documentation's example of this is about "engine" memory management, by which they mean the callbacks you provide to DLSS for it to be able to allocate VRAM. If you give it a callback that has suboptimal performance, or just one with a sleep(5000) in it, then yes, that will affect the time to get your upscaled color buffer back, because it has to allocate some VRAM before it can do that. But that has nothing to do with the execution time of the actual upscaling operation, which, again, is what people are always talking about here.

I am not talking about the documentation's examples. I'm talking about the very fact that it makes a claim that directly contradict's your claim. Are you going to acknowledge that or keep acting like it doesn't exist?

 

[Selarmor]

No, that's why it's often said it's too early to say. However, if Xbox Series S do get GTA6, unless the game is severely CPU-bound, I think it'd be reasonable to say a Switch 2 port is possible.

When I said only runs at 30fps the implication was that it's CPU bound, though I guess they could choose not to have a performance mode even if it was possible.

 

[darthdiablo]

When I said only runs at 30fps the implication was that it's CPU bound, though I guess they could choose not to have a performance mode even if it was possible.

I'm not sure that's right? If things run at 30fps, the implication is that it's CPU-bound? Because being GPU-bound is a thing too..

 

[Serif]

A CPU bound game may need more nips and tucks to the specific entities that are clogging up threads. NS2 is in a better position with the octo-core CPU, but it will still be clocked lower than the stationary devices. I am wondering what sort of operations could potentially be offloaded to tensor cores, or if they'll be too busy doing the upscaling part.

 

[Deleted member 511]

Erm, the DLSS documentation is available on Github.

It's also not supposed to be shared anywhere else.

 

[Danny]

I think what could stop GTA 6 from being ported to Switch 2 is if the game is only possible to port to Switch 2 by doing stuff like lowering number of NPCs and other stuff that reduces CPU load, because that would make the Switch 2 incapable of having online crossplay with other platforms and i don't think they would port the game to Switch 2 if the online play is only possible with other Switch 2 players.

 

[Softy]

I do wonder how NSO will operate on the Switch 2. They’ll obviously keep compatibility with the existing emulators on Switch and any games added on one of those should work on both Switch 1 and 2, but with the extra power and DLSS, is thinking the games will jump to output 4K that much of an ask? Plus rewind on N64 should be more feasible.


What are they saying? That they won’t say anything about the new Joy cons?

I'm curious about potential DS on NSO with the mouse con because ease of use depends on how easily you can switch to mouse con when in docked/tabletop mode along with where the mouse cursor decides to appear when you put down the joycon unless the cursor stays where you leave it and it only "touches" when you "click" which would probably make this better than the actual DS lol

 

[LiC]

I am not talking about the documentation's examples. I'm talking about the very fact that it makes a claim that directly contradict's your claim. Are you going to acknowledge that or keep acting like it doesn't exist?

It doesn't contradict me, as I showed by covering the documentation's one example of how the "engine" matters. Because I'm making a distinction between the total package, and the execution time of the DLSS upscaling operation. The latter is what people are always concerned with here, and what has (correctly) been described as being fixed with respect to output resolution. The reason this comes up so much is because of unfounded insecurities where they think DLSS might not be viable for Switch 2's use case, so people are always asking if there's some other knob you can turn to make it viable. The premise is wrong (it is in fact viable), but the answer is no, there's no performance knob besides the output resolution.

And on that note, having read more of the reply chain, I want to disclaim that I think Digital Foundry's unscientific DLSS test was badly conceived and its results irrelevant. And I wouldn't be caught dead using that to suggest DLSS for 4K 60 fps is impossible, which is ridiculous.

 

[bmac6502]

And the funny thing is that they will likely do as every other publisher, rush to make it Steam deck verified later, funny how all those seemingly impossible to port next gen AAA games all rush to steam deck verify their games afterwards.

That. GTA 6 is already known to be coming for Series S. We already know that despite being cut down further in brute strength, Switch 2 will have some advantages to offset, such as an extra 2 GB of RAM, and probably RT and Tensor acceleration that will still beat Series S because RDNA 2 was so weak in that regard and then cut a bit too deep on the Series S APU. The Switch 2 port challenge might be world density. Switch 2 will have as many cores, but lower clocks and lower storage speeds will make it hard. But that could be mitigated (as we’ve seen many games do before) with slightly lower geometric density and lower NPC density. Many games manage to make those cuts without impacting the experience too much.

Besides, even though PC ports usually come later from Rockstar, anyone who thinks they haven’t already planned for Steam Deck verified on GTA 6 is a fool, and if Steam Deck can do it, I think we can safely assume that puts Switch 2 within reach, even if Switch 2 comes in on the lower side of expectations.

Nintendo has sort of gotten what they wanted here. The Switch has made “AAA capable handheld” a viable category that didn’t exist in any meaningful form before. Valve followed that concept and now others have rushed in behind that. The result of that is that “AAA capable handheld” is now a valid target catagory that devs need to plan for when making games. That was not at all a thing when Switch launched, and now it is, which virtually guarantees most games in development are going to have some plan to get down to that range of hardware. That’s a permanent shift in the game buyer market that will affect dev decision making and absolutely plays to Nintendo’s advantage. Before when Nintendo released hardware a half or whole step behind, most devs could just go “well I guess we will just write off Nintendo ports” but PC handhelds mean all multiplats kind of need to have that plan with or without Nintendo, and that makes Nintendo ports more likely.

 

[ScribbleThread_]

I'm throwing my prediction into the ring:

I don't think GTA6 will come to the Switch 2, my main reason for having this opinion is portable mode the CPU clocks. The current gen home consoles will presumably be stuck at 30fps with exponentially more powerful CPUs compared to what the Switch 2 has on tap in portable mode.

We're expecting a 6-10W (estimated) 8nm portable console to run GTA6 when the PS5 Pro and XSX will be stuck at 30? Naw not happening. There's only so much you can do to lessen the CPU burden. This isn't something that can be rescued by having more VRAM than the Xbox Series S.

Another reason I don't expect it to come to Switch 2 is GTA Online. That's even more CPU-taxing than the singleplayer mode. I don't expect Rockstar to bring GTA6 to Switch 2 if they can't get online working

I edited my post, I forgot docked CPU clockspeeds are just as nerfed as portable to keep easy compatibility for devs

 

[Concernt]

I just don't think we have even close to enough information to discern that a game like GTA5 can be "easily" ported.

The future will tell.

Again, this is a speculation thread, responding to speculation with this doesn't keep to the topic. People are speculating that it will be easy to get working. People are free to discuss why they think it might not be, but this quite literally isn't the place for "the future will tell". We discuss what the future could hold based on the information we have.

I'd also like to remind everyone that this is primarily the hardware thread and not a software discussion thread. It's fine to discuss the technological barriers ( or lack thereof ) for given features and games, but primarily discussing the presence or absence of one game is not on topic.

 

[BDGAME]

Is this the correct translation?

If that is correct, then nintendo will have a big trouble.

Imagine if these joy-con are adapt to work on PC. So, before Nintendo reveal their new console, people will be showing playing joy-con + mouse-con on PC.

Right now, the wait is became more risky for Nintendo than ever.

 

[gabeplays]

That. GTA 6 is already known to be coming for Series S. We already know that despite being cut down further in brute strength, Switch 2 will have some advantages to offset, such as an extra 2 GB of RAM, and probably RT and Tensor acceleration that will still beat Series S because RDNA 2 was so weak in that regard and then cut a bit too deep on the Series S APU. The Switch 2 port challenge might be world density. Switch 2 will have as many cores, but lower clocks and lower storage speeds will make it hard. But that could be mitigated (as we’ve seen many games do before) with slightly lower geometric density and lower NPC density. Many games manage to make those cuts without impacting the experience too much.

Besides, even though PC ports usually come later from Rockstar, anyone who thinks they haven’t already planned for Steam Deck verified on GTA 6 is a fool, and if Steam Deck can do it, I think we can safely assume that puts Switch 2 within reach, even if Switch 2 comes in on the lower side of expectations.

Nintendo has sort of gotten what they wanted here. The Switch has made “AAA capable handheld” a viable category that didn’t exist in any meaningful form before. Valve followed that concept and now others have rushed in behind that. The result of that is that “AAA capable handheld” is now a valid target catagory that devs need to plan for when making games. That was not at all a thing when Switch launched, and now it is, which virtually guarantees most games in development are going to have some plan to get down to that range of hardware. That’s a permanent shift in the game buyer market that will affect dev decision making and absolutely plays to Nintendo’s advantage. Before when Nintendo released hardware a half or whole step behind, most devs could just go “well I guess we will just write off Nintendo ports” but PC handhelds mean all multiplats kind of need to have that plan with or without Nintendo, and that makes Nintendo ports more likely.

Very well said

 

[Softy]

I just want to the ports were done by Nvidia Lightspeed studios, which no doubt contributed to their excellent performance. It definitely wasn't a quick and dirty port.

That being said, Im kind of puzzled why they never released HL2 and its episodes, as I assume a large part of the engine work is already done.

Maybe they knew about half life 2 RTX + 20th anniversary plans and decided it would serve better as a switch 2 launch port since it would be able to run way better and show off stuff that is only possible now with the switch 2 hardware?

Edited to make it more inline with hardware speculation

 

[roleparadise]

Wii U has 176GFLOPs and runs the game at 720p30fps, if we double that to 352GFLOPs to get the raw performance for 60fps, lets actually make it 380GFLOPs so it can be a stable 60FPS right? Now raw performance needed to go to 4K60fps is 3.4TFLOPs, that is without any optimization for a GPU architecture about 12 years newer than Wii U's... This is overkill for 4K60fps botw. It requires no upscaling, and botw has it's vector data exposed iirc, so TAA type of upscalers such a DLSS would work on the device as well.

Trying to make sure I understand here. Are you suggesting that the Switch 2 GPU docked is capable of ~11.5x the performance as Switch 1 docked (900p30 -> 4k60 native)? Or I suppose even more when accounting for architecture. Certainly seems high, but exciting if true. I was thinking more like 7-8x based on what others were saying.

Also, is it an accurate measure of practical performance increases to compare raw flops here? Or maybe that is why you took architecture factors out of the equation, to illustrate a baseline?

 

[CloudOccasion]

Maybe they knew about half life 2 RTX + 20th anniversary plans and decided it would serve better as a switch 2 launch port?

Can it reasonably do it with path tracing? Portal RTX ran at 40~45fps on a 8GB RTX 3050 with 1080p DLSS performance (540p) with 4.7GB VRAM allocation (it didn't run out of VRAM).

 

[Deleted member 511]

It doesn't contradict me

It explicitly contradicts this claim:

The engine/pipeline and input buffers don't matter at all to DLSS execution time.

You've since clarified that you were talking about the upscaling operation time, but that is not interchangeable with DLSS execution time. If you cannot hold yourself accountable for this mistake, it is not a good look.

Furthermore, by definition, an algorithm cannot upscale from an input in a vacuum. The algorithm doesn't have inputs, it receives them externally and will always be dependent on that external integration, which will affect the times during which it could do its job. Even in the test example used without a 3D renderer, it still required integration with the command line to measure execution times.

Asking what an upscaling operations' cost is without factoring in integration is a curious endeavor but ultimately constrained by the fact that it's a moving target, changing with changes to hardware and algorithmic improvements.

EDIT:

Forgot to mention the quantization to 8 bits called out by NVIDIA when the documentation covers LDR/HDR. This is talking about the upscaling operation, so yet something else besides the output resolution that affects its performance directly (but also indirectly because it needs good tone mapping/perceptually linear encoding to work properly).

 

[Concernt]

Can it reasonably do it with path tracing? Portal RTX ran at 40~45fps on a 8GB RTX 3050 with 1080p DLSS performance (540p) with 4.7GB VRAM allocation (it didn't run out of VRAM).

I think with specific effort, pathtracing should be on the table. Even if it means resolutions go lower and a 30FPS cap, this system has plenty of memory for its performance level (important for raytracing, as I understand it) and is designed around being, well, an RTX console. Nvidia Lightspeed working with their own technology should be able to get things running, even if there's cuts like fewer rays per pixel rendered, heavy handed ray reconstruction, etc.

 

[Dimensio]

The current gen home consoles will presumably be stuck at 30fps with exponentially more powerful CPUs compared to what the Switch 2 has on tap in portable mode.

The CPU doesn't change between docked and handled mode.

 

[Enryu]

I'm throwing my prediction into the ring:

I don't think GTA6 will come to the Switch 2, my main reason for having this opinion is portable mode. The current gen home consoles will presumably be stuck at 30fps with exponentially more powerful CPUs compared to what the Switch 2 has on tap in portable mode.

We're expecting a 6-10W (estimated) 8nm portable console to run GTA6 when the PS5 Pro and XSX will be stuck at 30? Naw not happening. There's only so much you can do to lessen the CPU burden. This isn't something that can be rescued by having more VRAM than the Xbox Series S.

Another reason I don't expect it to come to Switch 2 is GTA Online. That's even more CPU-taxing than the singleplayer mode. I don't expect Rockstar to bring GTA6 to Switch 2 if they can't get online working

My eyes almost rolled inside my head.

 

[Tentacle-tropes]

I'm throwing my prediction into the ring:

I don't think GTA6 will come to the Switch 2, my main reason for having this opinion is portable mode. The current gen home consoles will presumably be stuck at 30fps with exponentially more powerful CPUs compared to what the Switch 2 has on tap in portable mode.

We're expecting a 6-10W (estimated) 8nm portable console to run GTA6 when the PS5 Pro and XSX will be stuck at 30? Naw not happening. There's only so much you can do to lessen the CPU burden. This isn't something that can be rescued by having more VRAM than the Xbox Series S.

Another reason I don't expect it to come to Switch 2 is GTA Online. That's even more CPU-taxing than the singleplayer mode. I don't expect Rockstar to bring GTA6 to Switch 2 if they can't get online working

You’d be right that they won’t bring over GTA6 if GTAO isn’t included. Though I doubt power will be the main reason preventing. Hearsay back in the day laid the problem at Nintendo’s online capabilities, though in what capacity was never really specified. As we’ve seen this gen, power is about the last reason for a majority of devs for why they didn’t bring games over.

People vastly overstate the power issue as an actual issue preventing ports. I keep going back to the Ace Combat example:

• 2017: Ace Combat devs say that Switch would be “most likely unable to handle their engine”
• 2022: admit that they did no form of assessment until someone randomly talked about doing a port. Found it was completely viable though had to farm it out because they had no expertise. They made a bad assumption w/o verifying if they were true.

If a company wants a game on a system they’ll get it on the system. Games like Ark & MK1 show this even if the initial ports are rough.

 

[Enryu]

Porting Witcher 3 to Switch took a year. I woudnt say that's "easily".

I would imagine similar for good Switch 2 ports.

1 year sounds average. Also, that's time, not difficulty. Also also switch 2 will be closer to ps5 than switch was to ps4.

 

[Future Boy Nemo]

The current gen home consoles will presumably be stuck at 30fps with exponentially more powerful CPUs compared to what the Switch 2 has on tap in portable mode.

Depending on what Nintendo clocks this thing at, PS5's CPU will be 2-3 times faster than Switch 2. Will turning down crowd density and other adjustments be enough to allow S2 to run GTA VI? I don't know, but it's not an exponential difference in power.

 

[abzinth91]

Yeah... no.

That's not even remotely comparable. It wasn't Sony's vision to have the weakest hardware with the ps2, or have bottlenecks with the ps1 etc. Poor design does not equal vision. Nintendo made the Gameboy based on Yokoi's philosophy. They made the Wii based on Iwata's philisophy. Not because they were incompetent hacks.

Sony's main vision throughout the years has been to see what the other guys do, and then double it. Nintendo had 2 shoulder buttons for the SNES, they doubled them for their first PS1 controller. Nintendo released a controller with a thumbstick, Sony revised their controller and added 2. Nintendo and Sega (iirc) released rumble, Sony added 2 rumble motors.

Yeah, I can remember (as someone who had a joystick on his PC) how cool the addition of analogue sticks for home consoles was. I replayed every PS1 game I had, after I got a Dual Analog (was that the name?)

 

[BDGAME]

Can someone wayyyy smarter than me explain in nerdy detail why the PS4 comparison is dumb and how good ampere is for the system in a way that will make me excited?

Ps4 is a unreal engine 4 hardware made for run 1080p native games.

Switch 2 is made for run Unreal engine 5 games, with all that engine can offer, but never on native resolutions. It's a DLSS hardware.

Basically, if ps4 run a game at 1080p60fps and they try to port it to switch 2 using the brute force, maybe it can run worse than ps4 on portable (less resolution or unstable frame). They need to adapt it to use switch 2 way of work to really make a good port. That way they can make a even better version than the PS4 had.

Switch 2 has advantages on better CPU, with more RAM and fast too. That can bring games with more phisics, more FPS and less or none load time.

So, compare these 2, even if they both have 1.8 TFlops (portable vs Ps4) is simplify too much what switch 2 will be.

 

[alucard0712]

I'm throwing my prediction into the ring:

I don't think GTA6 will come to the Switch 2, my main reason for having this opinion is portable mode. The current gen home consoles will presumably be stuck at 30fps with exponentially more powerful CPUs compared to what the Switch 2 has on tap in portable mode.

We're expecting a 6-10W (estimated) 8nm portable console to run GTA6 when the PS5 Pro and XSX will be stuck at 30? Naw not happening. There's only so much you can do to lessen the CPU burden. This isn't something that can be rescued by having more VRAM than the Xbox Series S.

Another reason I don't expect it to come to Switch 2 is GTA Online. That's even more CPU-taxing than the singleplayer mode. I don't expect Rockstar to bring GTA6 to Switch 2 if they can't get online working

I think something like 540p then upscale with DLSS is completely possible. Disabling ray-tracing (at least in portable mode) will help a lot too.
Another thing is - it's Rockstar. They made GTA5 run on a PS3. Not impossible to say the least.

 

[thoughthaver]

That. GTA 6 is already known to be coming for Series S. We already know that despite being cut down further in brute strength, Switch 2 will have some advantages to offset, such as an extra 2 GB of RAM, and probably RT and Tensor acceleration that will still beat Series S because RDNA 2 was so weak in that regard and then cut a bit too deep on the Series S APU. The Switch 2 port challenge might be world density. Switch 2 will have as many cores, but lower clocks and lower storage speeds will make it hard. But that could be mitigated (as we’ve seen many games do before) with slightly lower geometric density and lower NPC density. Many games manage to make those cuts without impacting the experience too much.

Besides, even though PC ports usually come later from Rockstar, anyone who thinks they haven’t already planned for Steam Deck verified on GTA 6 is a fool, and if Steam Deck can do it, I think we can safely assume that puts Switch 2 within reach, even if Switch 2 comes in on the lower side of expectations.

Nintendo has sort of gotten what they wanted here. The Switch has made “AAA capable handheld” a viable category that didn’t exist in any meaningful form before. Valve followed that concept and now others have rushed in behind that. The result of that is that “AAA capable handheld” is now a valid target catagory that devs need to plan for when making games. That was not at all a thing when Switch launched, and now it is, which virtually guarantees most games in development are going to have some plan to get down to that range of hardware. That’s a permanent shift in the game buyer market that will affect dev decision making and absolutely plays to Nintendo’s advantage. Before when Nintendo released hardware a half or whole step behind, most devs could just go “well I guess we will just write off Nintendo ports” but PC handhelds mean all multiplats kind of need to have that plan with or without Nintendo, and that makes Nintendo ports more likely.

it's a bit funny that there's only at best like 3 million steam decks but everyone is rushing to make it steam deck verified. bodes well for the succ tho.

 

[Eric_S]

GTA6 should be the biggest seller on the market, yes? Shouldn't it be of interest of Rockstar to have it run on quite a wide array of devices and specs?

Seemingly, a rather large swath of the PC market doesn't have 8 cores or more for example: https://store.steampowered.com/hwsurvey

My guess from afar, is that of GTA6 doesn't come for the Switch 2, it's something else that is hindering them. Like some odd quirk of its online capabilities or simply corporate politics.

 

[fwd-bwd]

Embeds for twitter have been disabled. Please refer to our site rules for more information.
Original post: https://tw*tter.com/tweet/status/1878452018668519776

Regular prototype buyer/owner

My Japanese reading is quite rudimentary, but I’m not sure about the machine translation saying that Joy-Cons leaked. It reads to me more like: “There are too many Switch 2 Joy-Con leaks”, referring to the leaked info not physical product (which would’ve been a serious crime). The machine translation for the next sentence is fine, but again we don’t know if “my friend might have it” is referring to info or product.

On a related note, some of you may remember the notorious Apple leaker Kang (who’s an executive of an accessory maker) posted a few vague messages about Switch 2 last November. He just posted a new message:

In recent days the real console is available for trying on accessories. I dare not try.

It looks like the real product is indeed available for people with means to examine. But it must be under the table, therefore Kang is reluctant to do it. So Genki’s CEO probably wasn’t lying about having access to a real console. (He later retracted that statement, but we know why.)

 

[ScribbleThread_]

The CPU doesn't change between docked and handled mode.

Oops my bad I forgot they scaled back clocks in docked mode to match portable clocks

For some reason I thought the CPU clocks were boosted over portable when docked

Let me edit my post

 

[LiC]

You've since clarified that you were talking about the upscaling operation time, but that is not interchangeable with DLSS execution time. If you cannot hold yourself accountable for this mistake, it is not a good look.

I didn't make a mistake. You quoted a sentence saying that DLSS always takes the same amount of time to upscale an image to 4K, which is true, by saying that it isn't fixed and is affected by the engine/pipeline, which is untrue. That was all the context you posted, so that's all I responded to. However, like I said, after reading the original posts and not just the single sentence you quoted, I see now that the discussion originated in Digital Foundry's misinformation, where their test was affected by many other factors besides just the upscaling execution time (and that part is also suspect since a 2050 or whatever they used isn't GA10F) and was worthless for measuring that execution time or drawing any conclusions about DLSS in general.

 

[JPQG]

Depending on what Nintendo clocks this thing at, PS5's CPU will be 2-3 times faster than Switch 2. Will turning down crowd density and other adjustments be enough to allow S2 to run GTA VI? I don't know, but it's not an exponential difference in power.

That is a very pessimistic view on the clocks. The end result may be way better than that.

 

[Hermii]

That is a very pessimistic view on the clocks. The end result may way better than that.

What do you expect the clocks to be? When we expected 4nm I would have said close to 2 GHz, but now I would say 1,5 GHz is optimistic.

 

[Deleted member 511]

I didn't make a mistake. You quoted a sentence saying that DLSS always takes the same amount of time to upscale an image to 4K, which is true, by saying that it isn't fixed and is affected by the engine/pipeline, which is untrue. That was all the context you posted, so that's all I responded to. However, like I said, after reading the original posts and not just the single sentence you quoted, I see now that the discussion originated in Digital Foundry's misinformation, where their test was affected by many other factors besides just the upscaling execution time (and that part is also suspect since a 2050 or whatever they used isn't GA10F) and was worthless for measuring that execution time or drawing any conclusions about DLSS in general.

So now you've resorted to a blatant lie. I don't have time for this.

Anyone who can access the DLSS documentation on GitHub, please reply to my post confirming that you have read the very first sentence under the DLSS execution time section and confirm whether it contradicts LiC's claim that DLSS execution time is not affected by engine integration or anything other than output resolution. Do not post the language verbatim, just reply with a response indicating whether that specific claim has been contradicted.

After this has been confirmed by at least 2 or 3 other members of this community, the evidence should stand on its own. I'm done with this conversation.

 

[Selarmor]

Can we stop with the weaponized pedantry over the colloquial use of the term DLSS? We all understand the context now.

What do you expect the clocks to be? When we expected 4nm I would have said close to 2 GHz, but now I would say 1,5 GHz is optimistic.

Do we have single core benchmarks for XSS Zen2 vs A78 at various clocks?

 

[Deleted member 735]

If this is an attempt to bait me into posting the documentation here, it's not going to work.

Why is your whole thing constantly posting about how you’re privy to “secrets” you don’t want to talk about? It’s giving huge “guy who graduated high school and is still hanging out here for some reason” energy, and I find that to be a very frustrating contribution to the thread.

There’s so many kind, technically knowledgeable people in here including the poster you are replying to. You continously post sanctimoniously implying that you know more than them, but never actually want to discuss details. Why are you here, then? Surely there are other people who’ve signed the NDAs you could talk to.

We’ve had extensive discussions about the DLSS documentation, the computational cost, the architecture in here - it’s not a mystery. (And even if it wasn’t public, it takes all of five minutes to sign up for the Nvidia developer program).

 

[Hermii]

Anyone who can access the DLSS documentation on GitHub, please reply to my post confirming that you have read the very first sentence under the DLSS execution time section and confirm whether it contradicts LiC's claim that DLSS execution time is not affected by engine integration or anything other than output resolution.

Can confirm.

 

[Deleted member 511]

Why is your whole thing constantly posting about how you’re privy to “secrets” you don’t want to talk about?

This is complete bullshit.

 

[bmac6502]

Trying to make sure I understand here. Are you suggesting that the Switch 2 GPU docked is capable of ~11.5x the performance as Switch 1 docked (900p30 -> 4k60 native)? Or I suppose even more when accounting for architecture. Certainly seems high, but exciting if true. I was thinking more like 7-8x based on what others were saying.

Also, is it an accurate measure of practical performance increases to compare raw flops here? Or maybe that is why you took architecture factors out of the equation, to illustrate a baseline?

Doesn’t seem crazy to me. Let’s do a big hand wavy over architectural and theoretical max efficiencies here, and just compare cores and clocks. We know the cores, we don’t know the clocks. The cores is 1536 for T239 and 256 for X1. 1536/256 is exactly 6, which means, (again hand waving over a lot of other factors that could cause non-linear scaling) you are halfway there on core counts alone. To get 8x you need a 33% clock increase over Switch 1 clocks. 11.5x on paper needs close to a 2x clock bump which seems high, but again that’s before considering efficiency advantages, RT cores, Tensor cores, and other new features that offer better/faster ways to handle certain workloads.

 

[LiC]

So now you've resorted to a blatant lie. I don't have time for this.

Anyone who can access the DLSS documentation on GitHub, please reply to my post confirming that you have read the very first sentence under the DLSS execution time section and confirm whether it contradicts LiC's claim that DLSS execution time is not affected by engine integration or anything other than output resolution. Do not post the language verbatim, just reply with a response indicating whether that specific claim has been contradicted.

After this has been confirmed by at least 2 or 3 other members of this community, the evidence should stand on its own. I'm done with this conversation.

You've totally lost the plot. So I agree there's no point in continuing the conversation.

One thing I do want to clear up for others reading this though, I think you're still confused about the confidentiality of DLSS documentation. It's right here in a public Github repo. You don't need any kind of account or credentials to access it.

 

[ShadowFox08]

Not to detract from your point, but how would they cool this?

More games is better to add a large pool of players, yeah. I’m mostly curious on CoD. It’s set to come to the system (through agreements), but on the technical side I wonder what unique Nvidia optimizations they perform. COD isn’t optimized for Nvidia like it is AMD being done externally, but if it’s done internally and with the right care that will translate to PC.

I didn't see the TFLOPs error until you pointed it out clearly. lol

That would leave no room for graphical improvements though like ray tracing and such

I do think there's a decent chance we'll get ray trace support to show the capabilities of it on Switch 2. Will be interesting to see multiple modes by developers (performance, RT mode, etc) For handheld, I'm thinking DLSS will be used. 720p native and then upscaled to 1080p at 30 or 60fps with RT.

 

[gabeplays]

No, that's why it's often said it's too early to say. However, if Xbox Series S do get GTA6, unless the game is severely CPU-bound, I think it'd be reasonable to say a Switch 2 port is possible.

It's just as fruitless of a discussion as is debating, say, what the node is. We simply don't have enough information. It's a game that nobody has seen before on a console nobody has seen before, nevermind any games. There's hardly anything to work with. All we have are surface-level hypotheticals.

It's not that people can't speculate, and there can be very valid and articulate arguments, but ultimately there's a hard wall.

 

[Birdo The Great]

FamiBoards when they finally announce the node: